Lessons Learned Sciencing The Web
summary
Argues that the right specifics decide optimization: measure with BigQuery, WebPageTest, Lighthouse, Puppeteer and the Chrome User Experience Report, then optimise by code‑splitting/PRPL, preload, service workers and caching. Gives numbers: 10% of sites ship ~1 MB JS, 50% ship ≥300 KB; Pinterest cuts JS from 620 KB to 150 KB bundle and gets interactive in under six seconds (was 23s); time‑to‑interactive goal ~5s, Twitter reduces repeat‑visit loads 6s→1.5s; use Brotli, MozJPEG, image sizing, and monitor budgets.
AI-written summary · about
chapters (10)
- 0:00Start
- 5:08Can we data-science the web?
- 6:17MOTO REMY
- 7:29Queryable RUM for the web?
- 12:55Removing unused code can reduce network transmission times, CPU-intensive code parsing, and memory overhead
- 14:06LEARN FROM GAME DEVELOPERS
- 16:37INTRODUCE WORKFLOWS THAT FORCE EVERYBODY TO THINK ABOUT LOADING TIMES FROM THE BEGINNING.
- 31:51Have a Web Font Loading Strategy
- 39:07EVERYONE IS RESPONSIBLE FOR performance.
- 40:56MEASURE OPTIMIZE MONITOR
related talks
41mSpeed at Scale: Web Performance Tips and Tricks from the Trenches
37mThe Cost of JavaScript
36mNobody cares about web performance. It's hard, expensive and unneeded
35mRevitalising the Web - The Importance of Web Performance
24mSuperpowers of browser's Web API
28mA Case Study
29mSpeed Essentials: Key Techniques for Fast Websites
01h 03mSpeed up your Web Application